Radboud University Nijmegen, Faculty of Arts
Department of Economic, Social and Demographic History
From Class to Citizenship.
Global Inequality and Social Mobility in the Twentieth
Century
Master thesis
Date: 15 August 2016
Author: Elien van Dongen
Supervisors: Prof. dr. J. Kok, Ingrid van Dijk Msc
Acknowledgements
I could not resist the temptation to end the process of writing this thesis with some words of thanks, as I wouldn’t have been able to complete it without the moral support of family and friends. My first and dearest word of thanks goes to Dennis Lundström. Although I have not (yet!) been able to do the same for you, without your help and support, endless listening to my theories, practical help behind the computer, and most importantly hugs, I would not have been able to complete this thesis.
Secondly, I could not have done without the optimism and words of support from Annabelle Jansink, one of the best historians and writers that I know. If I would have had enough time to let you read my thesis, I am sure it would have been a delight to read. Similarly, the optimism and company of Evelyn Leiva Deantonia gave me the energy to write the final parts of this thesis. I also want to thank Chiara Fasotti, as there is nobody who understands the challenges that come with writing a thesis better.
My sisters, Veerle and Janne, deserve my gratefulness for all their practical support: without the dinners they cooked for me and the shopping they did for me this summer I would have surely aged five years by now. My parents deserve praise for this too, and I will certainly miss their everyday support.
I have to thank all the anonymous nerds on Stackoverflow. For every question I ever had about programming, an answer was only two clicks away. And finally, I want to thank my supervisors, Jan Kok and Ingrid van Dijk, who not only helped me with my thesis but also more generally supported me in my
academic pursuit. For this, I also want to thank Angelique Janssens, my supervisor in socioeconomic history more generally, who encourages me with her endless optimism.
Contents
Acknowledgements...2
Introduction...5
Chapter 1: Background...7
Inequality of opportunity in the age of globalization...10
Research question: migration and global inequality...11
Factor price equalization and capital labour substitution: an economic approach to migration and
economic convergence...13
The wealth and poverty of nations: why some are so rich and some so poor...15
Chapter 2: Data and Methodology...18
Global bilateral migration flows...18
Clustering of world-regions...18
Wages and incomes...20
Economic inequality: ginis...23
Remittances...24
Education...24
Urbanization...25
Internal and international armed conflict...25
Regression analysis...27
Chapter 3: Results...31
Model selection...31
General results...33
Emigration...36
Immigration...38
Other covariates...39
Conclusion and Discussion...41
Bibliography...44
Attachments...46
Attachment 2.1...46
Attachment 2.2...48
Attachment 2.3...50
Attachment 2.4...51
Attachment 2.5...54
Attachment 2.6...56
Attachment 3.1...62
Attachment 3.2...63
Attachment 3.3...67
Attachment 3.4...71
Attachment 3.5...75
Attachment 3.6...78
Attachment 3.7...81
Attachment 3.8...84
Attachment 3.9...86
Attachment 3.10...88
Attachment 3.11...90
Attachment 3.12...92
Attachment 3.13...94
Attachment 3.14...96
Attachment 3.15...99
Introduction
In recent years, the nineteenth century ‘Great Divergence’ has become a popular topic in economic history.1 This economic divergence between ‘the West’ and ‘the Rest’ continued during the twentieth century,
and especially in this period increases in global inequality were largely due to an increasing economic inequality between countries.2 Contrary to this global trend, national wealth and income inequalities in the
western world, as well as inequalities between countries in the Atlantic economy, decreased during the first half of the twentieth century.3 At the same time the economy of countries in ‘the Rest’ of the world lagged
behind in relative, and sometimes even in absolute, terms.4 Although historians often discuss the latter half of
the twentieth century in terms of globalization, the huge economic gap between the West and the Rest persisted.5 We saw a high inequality in which class mattered less and less, while the state issuing your
passport mattered more and more. As a consequence we can now explain 66-73% of a world citizen’s income by the county he or she lives in.6 This means that a large part of your future income is determined by the
country you are born in. Interestingly, this inequality determined by citizenship seems to be stronger for those at the bottom of national income distributions.7
To understand the impact of changes in the global economy on the individual level of world citizens, we must combine between- and within-country inequality. Global studies of between-country inequality often compare countries at the national level and are written in a context of globalization, factor price equalization and convergence. More detailed studies into the socioeconomic status of individuals, on the other hand, are often confined to one, or a small number of, (western) countries. While ‘social mobility’ and ‘inequality of opportunity’ play an important role in the latter, they are rarely referred to in the context of international inequality studies.
Not only the size of inequality, but especially the ‘stability’ of inequality matters for the opportunity of people to change their socioeconomic position. In national studies social mobility is often measured in the
1 K. Pomeranz, The Great Divergence. China, Europe, and the Making of the Modern World Economy (Princeton 2000).
2 F. Bourguignon and Ch. Morrisson, ”Inequality among world citizens: 1820-1992” The American Economic Review 92:4 (September 2002) 727-744.
3 Th. Piketty, Capital in the 21th century (Paris 2013), see figure I.1 for income inequality, USA, figure I.2 for wealth inequality, western Europe.
4 B. Milanovic, Worlds Apart. Measuring International and Global Inequality (Princeton 2005); idem, ”A short history of global inequality. The past two centuries” Explorations in Economic History 48:4 (December 2011) 494-506.
5 Whether overall global inequality increased or decreased during the last quarter of the twentieth century is debated, see for a comprehensive overview of the literature on the topic: S. Anad and P. Segal, ”What Do we Know about Global Income Inequality?” Journal of Economic Literature 46:1 (March 2008) 57-94.
6 B. Milanovic, ”Global inequality of opportunity. How much of our income is determined by where we live?” Review of Economics and Statistics 97:2 (May 2015) 452-460. Based on a linear regression explaining annual average household per capita income in $PPP by country dummies, income data from 2009, R 2 = 0.733 unweighted, R 2 = 0.657 population-weighted.
7 L. Pritchett, ”The Cliff at the Border” in: R. Kanbur and M. Spence, Equity and Growth in a Globalizing World (Washington DC 2006) 263-286; B. Milanovic, ”Global Inequality. From Class to Location, from Proletarians to Migrants” Global Policy 3:2 (May 2012) 125-134.
form of intergenerational changes in the socioeconomic level of occupations. However, upward and downward mobility in this study are conceptualized in a broader sense, with upward mobility meaning a positive change in position within the global income distribution. Given the high influence citizenship has on ones position within the global income distribution, a good example of upward mobility then becomes migration from a poorer to a richer country.8 The opportunity for upward mobility is then increased by
liberalization of international labour migration. In this light, the pre-1930 migration wave from Europe to the ‘New World’ is often viewed as one of the key reasons why the Atlantic economy converged, as for example by the American economists Timothy Hatton and Jeffrey Williamson.9 However, while the latter part of the
twentieth century in many respects was an era of globalization, it was not with regard to opportunities for international labour mobility. Generally, the opportunities for labour mobility have been restricted instead of liberalized.10
To understand the impact of restricted immigration on the global economy at large, and on the opportunities for upward mobility of world citizens, I will try to answer the question if and how migration
flows have affected global inequality over the last fifty years. Specifically, I will investigate the effect of
immigration and emigration on the wages or incomes of specific occupational groups within countries between 1960 and 2015. A better understanding of the relationship between migration and economic growth and inequality would be of interest to the academic as well as political debate. But an answer to this question is not only of economic interest; changing trends in global inequality in the past have been linked to political turmoil, as have changes in migration flows.11
An analysis of the relationship between migration flows and global inequality will be done by means of OLS regression. For three occupational groups and seven world-regions separate models will be created to analyze the relationship between migration to and from different world-regions and wage growth in these occupational groups. The results of this analysis are presented in chapter three. In order to do such an analysis, however, a dataset of unskilled, skilled and high-skilled global wages was created. The construction of this database is discussed in chapter two, as is the construction of seven world-regions based on income and migration characteristics. The first chapter starts with a discussion of the literature on global inequality and migration.
8 6] C.V. Zuccotti, H.G. Ganzeboom and A. Guveli, ”Has Migration Been Beneficial for Migrants and Their Children? Comparing Social Mobility of Turks in Western Europe, Turks in Turkey, and Western European” International Migration Review (November 2015). Web; J.E. Roemer, Equality of Opportunity (Harvard University Press 2000).
9 T.J. Hatton and J.G. Williamson, Global Migration and the World Economy. Two centuries of Policy and Performance (Cambridge US, London 2005).
10 M. Ruhs, The Price of Rights. Regulating International Labor Migration (New Jersey 2013); A.S. Timmer and J.G. Williamson, ”Immigration Policy Prior to the 1930s. Labor Markets, Policy Interactions, and Globalization Backlash” Population and Development Review 24:4 (December 1998) 739-771.
Chapter 1: Background
A hundred years ago the world was caught up in what we now call the Great War. Hopes and dreams of eternal progress and a world too sophisticated for war were shattered. Traditional political institutions collapsed, as did the wealth of the old establishment. The Great War signalled the beginning of a gradual change in wealth distribution in the western world which was further enhanced by the Great Depression and the Second World War. High national wealth and income inequalities which had seemed impregnable dissolved.12 A combination of misfortunes for the wealthy and a changing political environment led to the
rise of the 'middle class' and a considerable improvement in welfare for the working classes. Not only their economic situation became better, disadvantaged groups in the western world successfully fought for equity in other domains such as gender and race.
From a global perspective, however, this decreasing inequality in the western world does not tell the whole story of the global economy in the twentieth century, in which an ever increasing international economic inequality changed the meaning of poverty and wealth. The economic inequality between
countries, or 'Great Divergence', borne by the Industrial Revolution and the great leap forward of the western world during the nineteenth century, only accelerated its course during the first half of the twentieth century – despite war and depression. While 'absolute poverty' - living at subsistence levels - had been a global phenomenon during the late nineteenth and early twentieth century, which engaged workers from England to China alike, the meaning of 'being poor' now diverged along national borders. By the end of the twentieth century, to belong to the poorest 5% of Denmark's population was equivalent to being richer than 97-99% of the Chinese and Indian population - even at purchasing power parity.13
Global historians often discuss the twentieth century in terms of globalisation: national political institutions became connected through international and global organisations and trade agreements, private enterprises expanded their scale to become multinationals, and from the perspective of a world citizen the world became more accessible through lowered transport costs and new communication technologies. People as well as private and public institutions gathered the knowledge and economic means to move across national boundaries. Economic convergence between a group of countries after the Second World War, mostly in what we now refer to as the 'western' or 'developed' world, led to an optimistic belief among scholars and politicians in the west that this growth trajectory would be followed by the rest of the world as well.14 Hopeful signs came from Latin America, Northern Africa and the Soviet Union, economies that
seemed well on track to catch up with Europe and the Anglosphere.
12 This process is summarized empirically in Thomas Piketty, Capital in the 21th century (Paris 2013), figure I.1 (income - USA) and figure I.2 (wealth – western Europe).
13 Branko Milanovic, “Global Inequality. From Class to Location, from Proletarians to Migrants” Global Policy 3:2 (May 2012) Web. 127-128.
14 As in various interpretations of Simon Kutznets, “Economic Growth and Income Inequality” American Economic
Review 45 (March 1955) 1-28; Robert Solow, “A contribution to the theory of economic growth” Quarterly Journal of Economics 70:1 (February 1956) and; Paul Samuelson, “International Trade and the Equalisation of Factor
However, during the subsequent decades most of these areas fell behind. New hopeful signs from East Asia kept the dream alive, but on their own do not provide much reason to believe in a natural development towards global economic convergence.15 Until 1970 inequality between countries mostly
increased, and even during the last decades of the twentieth century the huge gap between 'the west' and the rest persisted.16 While only about one-fifth of global inequality could be subscribed to between-country
inequality in 1820, as much as four-fifth of global inequality was due to between-country inequality in 2002 (see figure 1).17
18
Global economic inequality, and especially a growing inequality between countries, is thus one of the major phenomena that characterize the twentieth century's global economy. We saw a high and increasing inequality in which class mattered less and less, while the state issuing your passport mattered more and more. As a consequence we can now ‘explain’ (in a regression sense) 66-73% of a person's income by
15 The converging regions are based on the pre-1950 and post-1950 'convergence club' in: S. Dowrich and J. Bradford DeLong, “Globalization and Convergence” in: Michael D. Bordo, Alan M. Taylor and Jeffrey G. Williamson (eds.),
Globalization in Historical Perspective (Chicago 2003) 199-203.
16 Whether overall global inequality increased or decreased during the last quarter of the twentieth century is debated, see for an overview of the literature on the topic: S. Anand and P. Segal, “What Do We Know about Global Income Inequality?” Journal of Economic Literature, 46: 1 (2008) 57-94; M. Roser, https://ourworldindata.org.
17 Branko Milanovic, “A short history of global inequality: the past two centuries' Explorations in Economic History 48:4 (December 2011) 494-506, 500. The results are slightly different for different inequality measures: the Gini coefficient ranges from 28% in 1820 to 85% in 2002 (using 2005 PPPs), while the Theil index ranges from 8% in 1820 to 66% in 2002 (figure 1), reaching its maximum in 1980 at 74% and decreasing afterwards due to increasing within-country inequality. These country inequality measures are population-weighted. The between-country data is based on Angus Maddison's 'Historical Statistics for the World Economy' (version 2004), newest version here. Global inequality data is based on François Bourguignon and Christian Morrisson, 'Inequality among world citizens: 1820-1992' The American Economic Review 92:4 (September 2002) 727-744.
18 P.H. Lindert and J.G. Williamson, “Does Globalization Make the World More Unequal” in: M.D. Bordo, A.M. Taylor and J.G. Williamson (eds.), Globalization in Historical Perspective (Chicago 2003) 227-275, 230.
knowing the country he or she lives in.19 This means that a large part of your future income, save for overall
growth of your country's economy or international migration, is determined by the country you are born in. Importantly, former World Bank economist Branko Milanovic shows that this inequality determined by citizenship is not the same for all citizens of a country, but holds more strongly for those at the bottom of national income distributions. An unskilled worker from Beijing, Delhi or Nairobi can improve his income by a factor 11 by relocating to New York or London, while an engineer would 'only' make 3 times as much after relocating.20 This international inequality – and especially its concentration at the level of unskilled
workers – represents a severe form of inequality of opportunity: a disparity that cannot be resolved through personal effort. As it is larger among groups that are lower educated, explanations based on human capital and productivity arguments become problematic. This is shown by Harvard economist Lant Pritchett, too, in his measurements of the “cliff at the [US] border” – his name for the international gaps in the wages of equivalent labour. Pritchett does not only measure this citizenship cliff for otherwise – including ethnically – equivalent workers, but also compares the gains from international mobility to a lifetime of benefits through an international development program. He finds that, for example, the gains of earning a US wage belonging to his level of productivity for a Bangladeshi would in four weeks compensate for a lifetime of microcredit.21
Much of the literature on economic inequality is focussed on either one of two forms of global inequality: between-country or within-country inequality. Global studies of between-country inequality usually compare countries at the national level and are written in a context of globalisation, factor price equalization and convergence. More detailed studies into the socioeconomic status of individuals, on the other hand, are often confined to one, or a small number of, (western) countries.22 While ‘social mobility’ and
‘inequality of opportunity’ play an important role in the latter, and are from various perspectives seen as undesirable within national boundaries, they are rarely referred to in the context of international inequality studies. To understand the impact of changes in the global economy on the individual level of world citizens, we must combine these two forms of inequality and study inequality between world citizens, as was most vividly propagated recently by Lant Pritchett and Branko Milanovic.23
19 B. Milanovic, “Global inequality of opportunity. How much of our income is determined by where we live?”
Review of Economics and Statistics 97:2 (May 2015) 452-460. Based on a linear regression explaining annual
average household per capita income in $PPP by country dummies, income data from 2009, R²=0.733 unweighted, R²=0.657 population-weighted.
20 This is due to the negative correlation between national income and economic inequality: higher income, in today’s world, is associated with lower economic inequality. There is no reason to expect causation here, but it warrants scepticism about the assertion that inequality stimulates growth. B. Milanovic, “Global inequality. From Class to Location, from Proletarians to Migrants” Global Policy 3:2 (May 2012) 125-134.
21 Pritchett, “The Cliff at the Border”.
22 Renowned, recent, examples include: Joseph Stiglitz, The Price of Inequality. How Today’s Divided Society
Endangers Our Future (New York 2012); Thomas Piketty, Capital in the 21th century (Paris 2013).
23 See for example: B. Milanovic, Worlds Apart: Measuring International and Global Inequality (Princeton 2005); L. Pritchett, “The Cliff at the Border” in: R. Kanbur and M. Spence, Equity and Growth in a Globalizing World (2006) 263-286.
Inequality of opportunity in the age of globalization
In particular, this thesis will focus on inequality of opportunity and 'social mobility' among world citizens. Not only the size of inequality, but especially the 'stability' of inequality matters for the opportunity of people to change their socioeconomic position. In recent studies 'social mobility' is often measured in the form of “career and generational changes in the socioeconomic levels of occupations” within one country, or a small group of countries.24 Along with this habit, the meaning of the concept has in practice become
confined to national borders. The opportunities for social mobility are then said to increase through for example equal access to education and an increased share of earned income (e.g. wages) as opposed to
owned income (e.g. inheritance).
However, upward and downward mobility in this study are conceptualised in a broader sense, with upward mobility meaning a positive change in position within the global income distribution, and downward mobility meaning a negative change of this form. Given the high influence citizenship has on ones position within the global income distribution, a good example of upward mobility then becomes migration from a poorer to a richer country. The opportunity for upwards mobility is then increased by liberalisation of international labour migration (e.g. the European Union's Schengen zone) and decreased by restrictions on labour migration (e.g. late twentieth century approach of the 'West' towards prospective economic migrants from the 'Rest').
As in the Encyclopaedia Britannica entry on social mobility: “Throughout history international migration has been an important factor in upward mobility. One instance may be seen in the 19th-century migration of members of the working and peasant classes from Europe to the United States.”25 The pre-1930
migration wave from Europe to the 'New World' is indeed often viewed as one of the key reasons why the Atlantic economy converged, as for example by the American economists Timothy Hatton and Jeffrey Williamson.26 According to their model, migration accounted for at least 70 percent of Atlantic
convergence.27 In this sense, globalisation during the latter half of the twentieth century bears little
resemblance to Atlantic integration; while the period in many respects was an era of globalisation, it was not with regard to opportunities for international labour mobility. Generally, the opportunities for labour mobility have been restricted instead of liberalised: both by restrictive policies in rich countries, and by the increasing number of independent countries in the world.28
24 “Social Mobility” Encyclopaedia Britannica, Web. 25 “Social Mobility” Encyclopaedia Britannica, Web.
26 T.J. Hatton and J.G. Williamson, Global Migration and the World Economy. Two Centuries of Policy and
Performance (Cambridge US, London 2005).
27 Williamson and Taylor tried to quantify the effect of migration on Atlantic convergence by developing a
counterfactual model assuming no migration. When fully integrated capital markets are assumed – that offset the effect of migration as the capital flows to the New World would have been smaller without migration – migration still accounted for 70 percent of Atlantic convergence. Under the assumption of completely disintegrated capital markets, migration accounted for 125 percent of convergence. As in reality capital markets were not fully integrated, the contribution of migration was probably larger than 70 percent. A.M. Taylor and J.G. Williamson, “Convergence in the age of mass migration”, European Review of Economic History 1:1 (April 1997) 27-63.
28 Pritchett, “The Cliff at the Border”, figure 11.1; Branko Milanovic, Global Inequality. A New Approach for the Age
'Upward' and 'downward' mobility are ambiguous concepts; we can define them in an absolute sense, based on income per capita with or without purchasing power corrections, or in a relative sense. A relative approach, with the nation-state as reference group, has been common in the literature. Thus, common approaches to social mobility suffer from twofold epistemological nationalism; both in choice of subject, and in choice of reference group. Especially in migration studies this approach is problematic; it is rather unlikely that in the event of international migration the reference group of the migrants – those who they compares their income to – suddenly shifts from the country of origin to the destination country. If and when, and under what circumstances, this happens, is a whole field of study in itself. In this study relative mobility of a national occupational group with respect to other groups in the same nation, as well as with respect to other nations, will be analysed.
Research question: migration and global inequality
To understand the impact of restricted immigration on the global economy at large, and on the opportunities for upward mobility of world citizens, I will try to answer the question if and how migration
flows have affected global inequality over the last fifty years. Specifically, I will investigate the effect of
immigration and emigration on the wages or incomes of specific occupational groups within countries between 1960 and 2015. Thus, instead of representing the entire world population, this study is concerned only with the global labour force – and the effect of international labour mobility on their wage incomes. Wage incomes are problematic in studies concerning earlier periods – as much of the remuneration for labour was not monetary, and in studies focussed on accurately capturing overall inequality in one society (for example with the Gini coefficient), as capital income is distributed more unevenly than labour income. However, in this study the focus is on the effect of international labour mobility on income growth. Therefore ideally only labourers would be studied. As we cannot directly find the total incomes of those global citizens who are active on the labour market, wages seem the most accurate approximation. From a classical economic perspective, we may expect labour mobility – leading to labour scarcity or abundancy – to have a direct effect on labour costs and thus wages. Studying the effect of migration on capital incomes would however be a useful addition for future studies.
A better understanding of the relationship between migration and economic growth of specific socioeconomic groups in different countries (inequality) would be of interest to the academic as well as public debate. But an answer to this question is not only of economic interest; changing trends in global inequality in the past have been linked to political turmoil, as have changes in migration flows. By knowing how these processes affect different groups economically, policy makers will be in a better position to avoid adverse effects and exploit the growth potential of migration.
To study the relationship between migration and global inequality, three broad occupational groups are created for each available country: unskilled workers, skilled workers and highly skilled workers. Social
mobility, in this thesis, is thus not measured directly with micro-level data, but approximated by using wage data for larger national occupational groups. Therefore, variations within such groups cannot be taken into account. Since our main objective is to analyse the effects of migration on national occupational groups at large, this is not necessarily a disadvantage – although micro-level studies on these relationships would form a valuable complement to the study conducted here. This is a data-driven but also a necessary restriction on the debt of the study; to create a comprehensible overview of global income mobility, we need to organize world citizens into a limited number of groups, around which the historical narrative can revolve.
The effect of migration on the income of these socioeconomic groups will be analysed by creating a regression model. Other relevant factors, such as conflict, remittances, education and urbanisation, will be taken into account and – importantly - migration rates for different world-regions will enter the model separately. These world-regions are created on the basis of international economic inequality and immigration patterns: each world-region contains countries with similar incomes (measured by GDP per capita at PPP) and has similar bilateral immigration flows.29 By distinguishing between migrant flows from
different world-regions, we take into account political, cultural and economic differences between these world-regions.
Separate explanatory models are created for the various world-regions, in which migration from all these world regions can have a different effect on wage development. Earlier studies into the relationship between migration and income or inequality usually create one model including all observations. They subsequently attempt to control for political and cultural differences by adding indicators to the analysis that should capture these differences. As it is hard to determine which indicators actually capture such
differences, a more straightforward way of taking them into account is creating separate units of analysis, which is done here through the clustered world-regions. An obvious drawback of this method is that it gives less clarity about the source of differences between these clusters. A major advantage, however, is that separate models allow variation to vary between clusters: if we expect the trends and effects of the
explanatory factors to be different in a different cultural, political and economic context, we should expect the remaining variation to be different as well. This assumption therefore leads to a more accurate
representation of trends within clusters.
Finally, by comparing the different occupational groups and different income groups from the first analysis conclusions can be drawn on the relationship between global inequality and ‘social mobility’ in the form of migration. Hereby we can see what, during the last 50 years, the effect of migration on economic inequality as well as growth has been. Although the results will depend on the choice of data, the main advantage of this analysis will be the spatial and temporal comparability: even if the available data is not ideal, similar data will be used for all countries in the analysis. Furthermore, instead of country aggregates and estimates, data from the International Labour Organisation (ILO) on national wages at the occupational level will be used.30 This makes it possible to look at the effect of migration on specific socioeconomic
groups within countries, as well as the aggregate effects on a global or regional level. 29 Clustering is done with the DBSCAN method in scikit learn (Python).
The focus of previous studies into the economic effects of migration has been on immigration into the western world, both because of data availability and because studies were often conducted by western policy institutes. In such studies often no significant effect of migration was found – although there are both examples of positive and of negative effects.31 However, to get a better idea of the effect of migration on the
global economy and find characteristics that are less context specific, we need to widen our geographical scope. Moreover, it is likely that the economic effects of migration on poor countries are (relatively) larger than the effects on rich countries – as the entire economy of poorer countries is smaller. This is especially true for remittances, which are usually high (as a share of GDP) in countries with opportunities to migrate to richer neighbouring countries (e.g. in 2013: Lesotho 20%, Tajikistan 49%, Nepal 29%, Haiti 21%, Lebanon 18%).32
Factor price equalization and capital labour substitution: an economic approach to
migration and economic convergence
Next to the societal and historical relevance of the relationship between migration and global inequality, studying this relationship empirically is also of major relevance to economic theory, especially in light of existing assumptions on the behaviour of different factors of production. Based on factor price theory, it is traditionally assumed that migration – just like liberalisation of other factors, by for example free trade agreements – leads to economic convergence.33 More precisely, we assume ‘old-style’ inequality to rise
in the receiving country, and this same form of inequality to decrease in the sending country. That is, inequality between those earning a capital income (e.g. landowners, shareholders) and those earning a wage through labour will increase in the area where labour becomes more abundant, while capital becomes more scarce; the receiving country. Conversely, in the sending country we assume that the value of capital (which becomes less scarce) will decrease, while the value of labour (which becomes scarcer due to emigration) increases. Under the assumption of rational agents living in a free world, we assume that wage earners will move from high inequality (and low wage) countries to low inequality (and high wage) countries. Thereby labour and capital will become distributed more evenly over the world or region in which migration takes place, and thus capital and labour will be valued more similarly, leading to convergence.34 This theory is not
30 In a standardized form, the ‘Occupational Wages around the World’ (OWW) Database: R.H. Oostendorp, “The Occupational Wages around the World (OWW) Database: Update for 1983-2008” NBER (May 2012).
31 See for an example of negative effects: G.J. Borjas, “The Economics of Immigration” Journal of Economic
Literature 32 (December 1994) 1667-1717; idem, “The Economic Benefits of Immigration” The Journal of Economic Perspectives 9:2 (Spring 1995) 3-22; idem, “Immigration and Globalization: A Review Essay” Journal of Economic Literature 53:4 (December 2015) 961-974.
32 D. Ratha et al., “Migration and Development Brief” World Bank 24 (April 2015).
33 That is, in turn, under the assumption of limited and constant capital labour substitution, cf. Heckscher-Ohlin model. 34 It should be noted that the model is generally applied to trade in goods, not migration. However, both are seen as factors of production, with the main difference that in the traditional model equalization is achieved by trade instead of migration. There are many problems with the assumption underlying the basic model, an important one not mentioned here is treated by Robert C. Allen in “Technology and the Great Divergence: Global Economic Development since 1820” Explorations in Economic History 49:1 (January 2012) 1-16. Allen shows how technological development was local instead of neutral, in the sense that subsequent technologies were only economically beneficial for countries with a high enough capital to labour ratio (the frontrunners that developed them) and could thus not successfully be
only subscribed to widely among neoclassical economist, but also by critics of traditional economic theory such as Thomas Piketty.35
As not all individuals act rationally, or primarily out of economic interest, all of the time, and the world is not free but restricted and controlled by various legislative and governing bodies, it is not self-evident that this theory would hold true empirically. Additionally, as the old distinction between capital-owners and wage-earners gradually disappeared in the post-1945 period, the result of factor price
equalization becomes less obvious.36 However, because capital is still largely in the hand of the income-rich
– and more unevenly distributed over the population than wages – the effects may still exist to some extent.37
In the case of the Atlantic economy, we saw that large migration flows have in the past led to economic convergence.
However, even the results Williamson and Hatton have obtained on the Atlantic economy in the period 1870-1913 are somewhat mixed. While the ‘New World’ on the receiving side of migration, and the ‘European periphery’ on the sending side, behaved as the above theory would predict, the results for the ‘European core’ are mixed.38 In the European periphery inequality decreased both at the bottom and at the
top of the income distribution; the gap between wages of urban workers and GDP per capita decreased (at the bottom), while at the same time the wage-rent ratio increased (at the top, or ‘old-style’ inequality). The reverse happened in the New World.39 In the European core, however, inequality at the bottom increased, and
the increase in wage-rent ratio may be due to other influences than migration, most importantly a changing composition of wealth (away from land rents) and increased foreign investment (in, amongst others, land in
implemented elsewhere. A perhaps more fundamental critique on the model is given by (amongst others) Thomas Piketty, who questions whether the so-called ‘elasticity’ between capital and labour is constant at about one. That is, whether one factor can be replaced by the other, but only to such an extent that the contribution of capital to entire income remains constant (as in the Cobb-Douglas production function). See: Piketty, Capital, 263-264; Loukas Karabarbounis and Brent Neiman, “The Global Decline of the Labor share” Quarterly Journal of Economics 129:1 (2013) 61-103; Michael Elsby, Bart Hobijn and Ayşegül Şahin, “The Decline oof US Labor Share” Brookings Papers
on Economic Activity (October 2013) Web. 35 Piketty, Capital, 643-644.
36 Thomas Piketty, Capital, chapter 8, figure 8.3 and 8.4.
37 Christoph Lakner and Anthony Atkinson, “Wages, Capital and Top Incomes: The Factor Income Composition of Top Incomes in the USA, 1960-2005” W eb (2013) Web.
38 ‘European periphery’ = Denmark, Finland, Norway, Sweden, Italy, Portugal, Spain, Austria, Ireland. Note that Eastern Europe is largely absent. ‘European core’ = Belgium, France, Germany, Great Britain, The Netherlands, Switzerland. ‘New World’ = Argentina, Australia, Canada, United States. Wage to income inequality was rising more rapidly in the US than elsewhere in the New World, while rent to wage inequality rose slower there than in other New World countries.
39 In the European periphery wages grew on average 1.73% per year, while GDP per worker hour grew 1.6% per year over the period 1870-1913. In the New World wages grew 1.14% per year, while GDP per worker hour grew 1.77% on average per year. In 1870-1913 wage-rent ratios increased by on average 2.32% per year for the European periphery, while decreasing 3.03% per year in the New World. See Hatton and Williamson, Global Migration and the World
the New World).40 These mixed results on the European core do not necessarily disqualify the theory, since
migration flows from this area were rather low as compared to those from the European periphery.
In light of the perception of poverty and wealth as relative, instead of absolute, phenomena, there are some further questions that we should answer, even if factor price theories hold true, to be able to say what labour mobility means for global inequality between world citizens. Firstly, is this convergence largely the result of economic decline in the receiving areas, or of economic growth in the sending areas; what happens to the global economy at large? As migration from poor to rich was the most common form of migration in the world during the latter half of the twentieth century, in this study this translates to the question whether the economic benefits of migration over the last fifty years have been larger for the ‘developing’ (sending) countries, than the economic decline in the ‘developed’ (receiving) countries. An important additional question, taking into account the heterogeneity within nations, is whether migration has reduced inequality more in the ‘developing’ (sending) countries, than that it increased inequality in the ‘developed’ (receiving) countries. All of these questions, of course, are based on the premise that there is a measurable effect of migration on economic growth, which is what we should find out first.
The wealth and poverty of nations: why some are so rich and some so poor
Over the past decades, an extensive body of literature has treated the question why certain parts of the world became – and remained – so much richer than others. To study the relationship between migration and global inequality, we need to take into account other factors that might have affected global inequality simultaneously. Three lines of reasoning in this respect are discussed here, but there exist many other
explanations ‘exogenous’ to economics, most notably from an environmental (as for instance in the works of Jared Diamond) or cultural (post-colonial or Weberian) perspective.
Inspired by institutional economics the concept of ‘conditional convergence’ has been gaining influence during the last decades. The concept is used to show that countries in the post-1945 period did converge economically if other conditions are held constant (‘ceteris paribus’). These supposedly exogenous 40 Between 1870 and 1913 wages of urban workers in the European core increased by on average 0.90% yearly, while yearly growth in GDP per worker hour was much higher at 1.46%, on the other side the wage to rent ratio increased here by on average 2.32% per year. The reason why comparison of this last measure between the New World and European core is problematic is derived from the work of Thomas Piketty. In Capital in the 21th century he shows that the value of land relative to national income decreased in both the European core and the New World (USA), but much less so in the US (from 25% to 20% of total wealth), followed by Germany (from 40 to 20% of total wealth), then France (from 40 to 15% of wealth) and lastly Britain (from 30 to 5% of total wealth). This, together with spiking foreign investment from the European core in the New World, make the wage to (land) rent measure problematic. The total value of land in the US was worth clos to 100% of national income during the entire period 1870-1910, with only slight variation over the years. Total wealth increased from 400 to 500% of national (yearly) income. In contrast, value of land became one-sixth of its 1870 value in Britain by 1910 as it decreased from double to less than one-third of national income, while total wealth was stable at around 700% of national (yearly) income. In France, too, the decrease in land value as proportional to national income was sharp; from 300% to 100% in the pre-WWI period. Here, too, total wealth only changed slightly from just over 700% to just under 700% of national income. The German developments were similar to Britain and France, with land value decreasing with almost 150% down from 275% of national income and total wealth decreasing slightly from 700 to about 650% of national income. Thomas Piketty, Capital in the 21th
conditions include all kinds of political and cultural factors, so that such theories of ‘conditional
convergence’ are quite meaningless for real-world prospects on economic development (i.e. controlled for suppressive regime, North and South Korea may be found to be converging during the last decades of the twentieth century). Moreover, there is a causality problem in this type of explanation. In for example the conceptualization of ‘conditional convergence’ by the American economist Robert Barro variables such as life expectancy, male secondary-school attendance and fertility rates – variables that are generally highly correlated with economic development – are controlled for.41 Thereby, the concept of economic progress is
deprived of all its welfare components and becomes largely empty. Looking at economic development at a given level of health and educational development makes international inequality into a chicken-and-egg-problem.
Another line of theory on global inequality is related to politics and specifically to openness to trade and democracy. In the work of the economists Steve Dowrick and Bradford DeLong ‘openness’ is defined by five conditions that, if a country fulfils at least one of them, make a country closed. These conditions refer mostly to trade barriers and other protectionist and monopolistic measures, but also to the existence of a large black market and a ‘socialist economic system’. This openness dummy explained some of the economic divergence in the period 1960-1980, but its validity seems ambiguous (why does socialism necessarily make a country closed?) and mostly applicable to the exact time period for which it was developed (Cold War) while not being generalizable. This is confirmed by Dowrick and DeLong themselves, as their ‘openness’ measure did a poor job in explaining economic divergence after the 1980s.42
Finally, global inequality is often related to technological development and differences in
productivity. People in ‘the West’ are then assumed to earn more than people in ‘the Rest’ because they are more productive; because they have better technologies, because they outsource all non-productive jobs, because they have more ‘human capital’ or even because they are more hard-working. And indeed, workers in ‘the West’ are often employed in more productive (i.e. generating more output per hour of labour) jobs than workers in ‘the Rest’; the capital to labour ratio has been much higher in the West throughout the twentieth century.43 But since international inequality is most severe at the level of unskilled workers, where
it cannot be explained by skill premiums or higher productivity, technological development is not a sufficient explanation for global inequality. In the work of Lant Pritchett, it is shown that controlled for ethnicity, education, age, gender, urban residence and other productivity differentials, wages disparities between the US and 42 developing countries are still substantial. The average person in any of these 42 countries can
multiply his or her income by 3.41 by moving to the US; they could earn PPP $10,000 more yearly.44
41 Steve Dowrich and J. Bradford DeLong, “Globalization and Convergence” in: Michael D. Bordo, Alan M. Taylor, and Jeffrey G. Willamson (eds.), Globalization in Historical Perspective (Chicago 2003) 191-226, 203-206.
42 Ibidem, 209-216.
43 Robert C. Allen, “Technology and the Great Divergence: Global Economic Development since 1820” Explorations
in Economic History 49:1 (January 2012) 1-16
44 Pritchett, “The cliff at the border”, 271-274. The other non-quantifiable productivity differentials are approximated by dividing each country’s wage difference with the US by the wage difference between the US and Puerto Rico (1.5), as labour migration has since long been liberalized between the US and Puerto Rico. Without this correction the 42 developing countries could on average multiply their wage by 5.11 by doing similar work in the US instead of at home.
Moreover, these results would likely be more dramatic if similar comparisons would be made between developing and northern and western European countries, as European countries have higher wages at the bottom of the income distribution than the US, which is the group that migrants from developing countries would usually be compared to due to their years of schooling.
Overall, the economic literature on global inequality is substantial, but popular theories often do not align well with the existing data. The debate would therefore benefit from more empirical studies (such as the one conducted here), new perspectives, and more data-driven theorizing.
In the following chapters I will first discuss the data and methodology used in my analysis, followed by results and a conclusion. In the results chapter, the relationships between wage growth at different skill levels, migration and other covariates will be discussed. The conclusion consists of a summary of these results in the form of global trends and answers to the specific questions raised in this chapter.
For some countries (Yemen, Nigeria) the wages would be more than ten times as high in the US as at home. Note that Pritchett’s analysis is only related to productivity and does not try to explain wage difference between countries in general. Pritchett therefore attempts to correct for all variables related to productivity, but not for other macroeconomic factors (such as scarcity of labour vs. scarcity of capital) in the countries that could also explain wage differences. Such variables, however, are irrelevant from the individual perspective and are assumed to be solvable by liberalisation of labour mobility.
Chapter 2: Data and Methodology
In this chapter the main data sources used and their role in this thesis will be outlined. First the the central elements of the analysis, migration flows and occupational wages, are described. An account and explanation of the ways in which it was adapted before use in the analysis is given. Subsequently the same is done for the various auxiliary indicators used in the regression model. After setting out its elements, the analytical tool in this analysis – a variation on the ordinary least squares (OLS) regression model – and its use here are described.
Global bilateral migration flows
New data on global bilateral migration flows, created by statistician Guy Abel, will be related to wage, income and wage and inequality data for over 180 countries during the period 1960-2015.45 A major
advantage of migration flows, as compared to migrant stocks, is that they suffer less from country specific measuring methods and conceptualization. Especially for an analysis spanning a large geographical area this is important, as there are large international differences in the time span for which previous migrants are considered migrants, leading to international comparability issues in migrant stock data. Migrant stock data is available from, among others, the United Nations Population Division and the World Bank. Guy Abel has combined these sources to estimate migration flows for 10 – and from 1990 onwards 5 – year periods during the past fifty years, adjusting migrant stock variations for demographical factors such as fertility and
mortality.
However, this is not the only advantage of this new dataset over traditional migrant stock data. Migrant flows are presented bilaterally instead of per country. This makes it possible to distinguish migration patterns by country of origin and destination. That is, whereas most sources of migrant stock data only provides information on the overall number of people entering or leaving a specific country, bilateral flows show which countries are similar with respect to the nationalities they receive and the countries to which their inhabitants move.
Clustering of world-regions
To be able to use these advantages, while not creating explanatory models with unreasonably many covariates (e.g. each bilateral flow), clusters of countries will be created based on national income and migration patterns. The exact construction of the clusters based on migration can be found in attachment 2.4. In short, eight migration clusters are constructed in an automated way by placing emphasize on the direction rather than the sum of all emigration or immigration flows in a country. Therefore, countries with emigration 45 G.J. Abel, “Estimating global migration flow tables using place of birth data” Demographic Research 28:28 (March
2013) 505-546. G.J. Abel and N. Sander, “Quantifying Global International Migration Flows” Science 343:6178 (March 2014) 1520-1522.
to (and immigration from) similar other countries will likely be clustered together. The size of each bilateral flow is still taken into account, but instead of on a continuous scale presented as a categorical variable. To further distinguish clusters based on national income countries are labelled ‘low’, ‘mid’ or ‘high’ income, and for each income category the available migration clusters become separate clusters. This could lead to
clusters, but in reality gives a total of 15 clusters: there are, for example, not eight but only three migration clusters including high-income countries. A more detailed description of the construction of the clusters combining migration patterns and national income can be found in attachment 2.7. The countries included in each cluster can also be found in attachment 2.7, while some general features of the clusters will be discussed in the results chapter.
Population-weighted average migration rates for each income cluster to and from a specific country are used as explanatory variables for the wage development in that specific country. That is, immigration and emigration per income cluster, giving a total of six migration covariates. Migration rates are not further differentiates as an initial analysis showed that differentiating more clusters as explanatory variables did not add explanatory power to the model. However, more clusters are differentiated to create separate explanatory models for: models are created separately, but with largely the same covariates, for seven clusters. These clusters are combinations of the earlier mentioned fifteen clusters that were found to have a similar relationship between wage growth and migration flows (overall emigration and immigration, only two covariates). Using more than seven different clusters would also make the sample size for certain clusters too small to be able to add extra covariates to the model and differentiate between types of migration.
In the regression analysis the log of migration rates per cluster plus one is used. This is done because the distribution of migration rates is heavily skewed towards zero - there are many country combinations for which bilateral flows are zero. Therefore, the unexplained variation in the regression also is larger around zero (heteroscedasticity) leading to less reliable results. By taking the log of migration rates extreme values become less extreme and therefore heteroscedasticity is reduced. The constant one is added to the rates for several reasons: in order to have the model not automatically predict zero wage growth if any migration variable is zero, to avoid extreme variation for migration rates between zero and one, and to not have minus infinity values in the analysis. This also makes interpretation of the coefficients easier: now zero migration rates are still zero in the log migration rates.
Moreover, it may be expected that migration does not have an immediate, but rather a lagged effect on wage growth: not all migrants immediately enter the labour market, and wages do not immediately adapt to a changing labour supply. Thus, in the regression model migration rates are used from a five-year period six years before the respective five-year wage period. This has the added benefit of diminishing causality issues.
Wages and incomes
Instead of country aggregates and estimates, where possible wage data from the International
Labour Organisation (ILO) as available on LABORSTA on national wages at the occupational level will be
used. The ILO October Inquiry contains yearly data on occupational wages for 159 occupations in 171 countries from 1983 onwards and is currently the most thorough international source on occupational wages. Using this data makes it possible to look at the effect of migration on specific occupational groups, as well as the aggregate effects on a national, regional or global level. The main advantage of these data is that similar data can be used for all countries in the analysis, especially at each year. Temporal differences in data sources will also be geographically consistent.46 As presented on LABORSTA and ILOSTAT, the October Inquiries
have various comparability issues. Therefore, a standardized version created by Remco Oostendorp and Richard Freeman is used for the years up to 2008; the Occupational Wages around the World (OWW) Database.47
From this data source, wages in contemporary local currencies (LCU) were used and where necessary converted to the local currency in 2008. The combined wage dataset is later adjusted using inflation at consumer prices to obtain wages in constant 2005 LCU, and subsequently divided by the 2005 PPP conversion factor to obtain wages in constant 2005 international dollars.48 However, this conversion
factor combined with consumer price indices (inflation) does often not give an internationally comparable indication of real wages at constant prices. The absolute values obtained can therefore not be considered as accurate. In this context, however, only real wage change over time is relevant, and for such change wage time-series at constant prices are far better comparable internationally than series at current prices: a growth rate of 6% with 3% inflation should not be treated in the same way as a growth rate of 6% with 9% inflation. The PPP conversion is only done to facilitate finding outliers in the data series, but is irrelevant for the analysis itself since a constant factor does not affect growth rates.
Wage and income indicators for three different occupational groups are created for as many countries and years as possible. However, the October Inquires contain few data for some world-regions, such as former Soviet states, and are only available from 1983 onwards. Data from other sources is therefore also included, and has first been standardized and was then converged onto the October Inquiries.49 Combining
46 See for consistency of the October Inquiry data and the definition of the three occupational groups based thereon attachment 2.5.
47 Database by R.b. Freeman and R.H. Oostendorp at www.nber.org/oww/. Documentation: R.C. Oostendorp, “The Occupational Wages around the World (OWW) Database: Update for 1983-2008” Web. (May 2012).
48 Consumer Price Index and PPP conversion factor for private consumption are mostly from the World Bank’s World Development Indicators, but where World Bank data is not available inflation at consumer prices is derived from the CIA World Factbook (historical series at indexmundi.com), and if none of the other sources is available the implicit price deflator from the UN is used. Full documentation of the various sources used per country-year is available upon request.
49 Standardizing the ILO dataset consisted of combining data from different sources (where certain sources such as labour force surveys were consistently preferred over other sources such as commercial surveys), with different types (e.g. including or excluding overtime pay), different time units (ranging from hourly to annual), different local currencies over time, and different versions of the classifications of occupational categories/ economic activities. Full documentation on how the dataset is standardized is available upon request. For definitions of the original ILO dataset see here.
the various datasets had to be done manually for most of the data, but was done using a linear regression where enough overlapping years were available. If few overlapping years were available one source was added to the other by using the ratio between the two sources for the closest overlapping year available. If no overlapping years were available between the data sources they were combined using a ‘neutral’ assumption on wage growth for the time gap between the different datasets: each occupational group’s wage was assumed to grow at the same rate as national GDP in current local currency. If all of these methods gave ‘unrealistic’ results combining the sources was done manually on a case-by-case basis.50 All wages from the
ILOSTAT dataset were appended to the previous series using GDP growth between 2008 and 2009.51
The ILO wage dataset for the years 1969-2008 contains wages per ‘economic activity’. These economic activities are well differentiated for manufacturing, so that each economic activity includes wages for a specific task. Non-manufacturing activities, however, are rather broad. To select specific economic activities from the ILO wage database which represent wages for unskilled and skilled workers, the mean and variance of the ratio between each economic activity and the general categories ‘manufacturing’ and ‘construction’ were assessed both over years and over countries. Economic activities were not included in an occupational category if they varied much between countries or years. Moreover, they were not included if they were available for too few country-years to create an (un)skilled worker that is comparable between countries and years.52
The first occupational group, unskilled worker, has the widest geographical and temporal coverage and is based on wages in specific manufacturing sectors (textiles and wood industries) from the ILO as available for the years 1969-2008, and occupational wages from the ILO October Inquiries for each year 1983-2015 - until 2008 from LABORSTA, from 2009 onwards from ILOSTAT. In the years 1983-2008 either the ILO or LABORSTA data were used depending on which wages were available for most years, to make the concept and change over time as consistent as possible. The ‘secondary’ data series was then added to the ‘primary’ series using one of the methodologies described above (regression, GDP growth or ratio).
The wage of an unskilled worker as based on the October Inquiries consists of 27 different occupational categories, mostly in basic services (e.g. waiter, grocery store re-stocker or postman) and manufacturing (unskilled labourers, packers and basic machine operators). For the years after 2008 the ILO has only made wages of general occupational categories available for public use, the wage of the unskilled worker is therefore equivalent here to the wages in the general category 'elementary occupations'. In
50 To perform any of the three operations described before for each series the original two series and the resulting series, if the operation would be applied, were printed. If these results contained outliers or irregular developments (e.g. a ten year period with every other year from a different data source giving alternating high wage growth and decline which was clearly due to data source differences instead of economic context) the adaptation was not accepted an an automatic note was created to adapt the series manually later. A full documentation of combining the datasets is available upon request.
51 GDP (growth) in current LCU from the World Bank World Development Indicators.
52 Full documentation on the selection process of categories included in the ‘unskilled worker’ and ‘skilled worker’ category based on ILO wage statistics is available upon request. The categories eventually selected can be found in attachment 2.1 and 2.2.
attachment 2.1 the exact definition of the 'unskilled worker' based on the wages in manufacturing and the October Inquiries can be found.
The second occupational group, skilled worker, is approximated by wages in the 'Transport, Storage and Communication' and 'Mining and Quarrying' sector from the ‘wages by economic activity’ database of the ILO, available in LABORSTA.53 Conceptually, this is not an ideal definition. However, the position of
these sectors within national income distributions is relatively stable over the period 1969-1990 and between countries, as compared to that of other sectors. This means that the sectors are likely to contain similar (or at least similarly paid) occupations for different countries, and thus represent the same economic category for different countries. The sectors both have somewhat higher wages than the overall manufacturing sector in almost all countries covered by ILO data, with wages in the sector 'Transport, Storage and Communication' on average 1.2 times those in the manufacturing sector and wages in the sector 'Mining and Quarrying' 1.35 times those in manufacturing over all countries and the period 1970-1990. They will include unskilled as well as skilled and high-skilled occupations. Furthermore, wages for these sectors are available for relatively many countries and years, which is of major importance for comparability. For later years these sectors are no longer appropriate to use for approximating the skilled worker's wage; the size as well as the wages of the mining and quarrying sector then decreases in most countries, while under the influence of new
communication technologies wages in that sector divert from those in the transport and storage sector. A better definition of the wage of a skilled worker is derived from the October Inquiries for each year between 1983 and 2008. It consists of 25 occupational categories in various sectors and includes skilled construction workers (e.g. carpenter, plumber, electrician), professional nurses, elementary school teachers, office clerks and stenographer-typists. When both sources are available, from 1983 onwards this data is always used and only supplemented by the economic activities data using regression, GDP growth or the closest ratio as before. From 2009 onwards wages of four major occupational categories from ILOSTAT are combined to form the skilled worker's wage.54 In attachment 2.2 the exact definition of the 'skilled worker'
based on the wages for general economic activities (only used for the years 1970-1980), the October Inquiries and ILOSTAT can be found.
The third occupational group, highly skilled workers, is only included for 1983 and subsequent years and is entirely based on the October Inquiries of ILO. For each year between 1983 and 2008 a combination of 26 occupational categories is used to create this group, including higher professionals (e.g. third level teacher, dentist, engineer, computer programmer) and executive officials in both the private and public sector. From 2009 it consists of technicians and other higher professionals, legislators, senior officials and managers. The exact definition of the 'highly skilled worker' for various years can be found in attachment 2.3.
53 Sectors based on the International Standard Industrial Classification of all Economic Activities (ISIC), revision 2 and 3 by the United Nations Statistics Division.
For the regression analysis log wage ratios over mostly (depending on data availability) five year periods are used. Log wage ratios are the logarithm of the percentage change between start and end year plus one, or equivalently the difference between log wage at the end year and log wage at the start year of a period:
If for a given start year the wage for the same skill level five years later is not available, instead the closest available year is used, with the restriction that the end year must be at least two years after the start year. By using changes over five year periods, instead of changes between years, short-term fluctuations have less effect on the eventual results. This choice is preferred over shorter periods as it is expected that migration will not be able to explain short-term fluctuations in a labour market, but rather have an effect on its gradual and long-term development. Note that ‘short term’ here would refer to one year, and therefore would not capture effects of seasonal migration. Moreover, we should not expect to capture effects of short term migration in the form of seasonal workers, as seasonal workers’ remuneration is probably not well represented in the used labour statistics (ILO wage database) and under-represented in the migration database as well. By creating a growth rates over longer periods with variable start and end, as opposed to for example using benchmark years (which may be represented by other years), as much of the available data as possible can be used, and periods are always exactly as long for all indicators: all covariates are adapted so that they represent change over the exact same period as wages.
For example, the USSR may have unskilled wages for 1969 until 1976, then have a gap until 1988. In this case 1969-1974, 1970-1975 etc. are included as five year periods, but only one variable, the log wage growth between 1976 and 1988, is created for the period without other data. All other indicators will also be taken for the period 1976-1988 and therefore be proportional to this longer period of wage growth.55 An
immigration indicator will include all immigration from a certain cluster to the USSR between 1976 and 1988, while the international conflict indicator will be the sum of all conflicts between 1976 and 1988, etc.
Economic inequality: ginis
One of the covariates which will be used to control the relationship between wage growth in a specific occupational category and migration is the Gini (national) inequality measure from the Standardized World Income Inequality Database (SWIID).56 It has the widest geographical and temporal coverage of
various inequality databases currently available, and is created with the purpose to improve international comparability of the Gini measure. During the period 1969-1985, it covers between 45 and 60 countries. For the subsequent period its coverage for the countries with occupational wages is nearly complete. Therefore,
55 It is assumed here that wage growth is exponential, as national income growth is usually assumed to be, and thus log wage growth should be linear with time. More on this in the description of the regression model.
56 Frederick Solt, “The Standardized World Income Inequality Database” Working paper SWIID Version 5.0 (October 2014). Web.
